A complete Deep Learning solution to predicting the behavior of a given cysteine. The predictions can be made either by using the features from the high resolution protein crystal structures or by using just the primary sequence data.
This is a fully functional web interface for the deep learning application built using the library Keras.
The code for this project is provided in a Github repository, making it convenient for other academics to replicate the results or even make further changes. Github Repository
The web interface is extremely intuitive and easy to use, the steps are outlined below.
Depending on what you prefer, choose "Structure Prediction" or "Sequence Prediction" button on the navigation bar, this will take you to the data collection form.
Fill in the parameters as requested by the form.
Once you have filled in the required blanks, click the submit button. The application takes approximately 10-15 seconds to predict. Please be patient.
The results page showcases the predictions made by the application. The structure based prediction model resulted in a macro-average of sensitivity of 79.25% on the test-1 dataset and 87.22% on the DUF dataset.